Genome-wide maps of UVA and UVB mutagenesis in yeast reveal distinct causative lesions and mutational strand asymmetries

Abstract

Ultraviolet (UV) light primarily causes C > T substitutions in lesion-forming dipyrimidine sequences. However, many of the key driver mutations in melanoma do not fit this canonical UV signature, but are instead caused by T > A, T > C, or C > A substitutions. To what extent exposure to the UVB or UVA spectrum of sunlight can induce these noncanonical mutation classes, and the molecular mechanism involved is unclear. Here, we repeatedly exposed wild-type or repair-deficient yeast (Saccharomyces cerevisiae) to UVB or UVA light and characterized the resulting mutations by whole genome sequencing. Our data indicate that UVB induces C > T and T > C substitutions in dipyrimidines, and T > A substitutions that are often associated with thymine–adenine (TA) sequences. All of these mutation classes are induced in nucleotide excision repair–deficient cells and show transcriptional strand asymmetry, suggesting they are caused by helix-distorting UV photoproducts. In contrast, UVA exposure induces orders of magnitude fewer mutations with a distinct mutation spectrum. UVA-induced mutations are elevated in Ogg1-deficient cells, and the resulting spectrum consists almost entirely of C > A/G > T mutations, indicating they are likely derived from oxidative guanine lesions. These mutations show replication asymmetry, with elevated G > T mutations on the leading strand, suggesting there is a strand bias in the removal or bypass of guanine lesions during replication. Finally, we develop a mutation reporter to show that UVA induces a G > T reversion mutation in yeast that mimics the oncogenic NRAS Q61K mutation in melanoma. Taken together, these findings indicate that UVA and UVB exposure can induce many of the noncanonical mutation classes that cause driver mutations in melanoma.

Introduction

The UV spectrum of sunlight is the primary etiologic agent for skin cancers such as melanoma (Garibyan and Fisher 2010), but the precise mechanism(s) by which it promotes melanomagenesis remain unclear. Genome sequencing of cutaneous melanomas and other skin cancers has revealed that nearly 90 % of somatic mutations in these tumors are C > T substitutions in dipyrimidine sequences (Bonilla et al. 2016; Hayward et al. 2017). This mutational signature can be largely explained by the fact that the primary UV-induced DNA lesions, cyclobutane pyrimidine dimers (CPDs) and 6-4 pyrimidine-pyrimidone photoproducts (6-4PPs), form exclusively at dipyrimidine sequences (i.e. TT, TC, CT, and CC; Friedberg et al. 2006). Both CPDs and 6-4PPs impede transcription and replication, and mutagenic bypass of CPDs during DNA replication primarily results in C > T or CC > TT tandem substitutions in dipyrimidine sequences, which comprise the canonical UV mutation signature (Pfeifer et al. 2005; Ikehata and Ono 2011; Brash 2015).

However, fewer than 50% of the driver mutations in melanoma fit this UV signature (Hodis et al. 2012). For example, BRAF V600E is the most recurrent oncogenic mutation in melanoma, but this is caused by T > A substitution that occurs in a nondipyrimidine sequence context (Thomas et al. 2006; Hodis et al. 2012; Hayward et al. 2017; Pandiani et al. 2017; Sample and He 2018). Similarly, the 3 most common driver mutations in the NRAS oncogene in melanoma (i.e. NRAS Q61R, Q61K, and Q61L) are caused by T > C, C > A, and T > A substitutions, respectively. To what extent these mutations are caused by solar UV exposure is unknown.

Genome sequencing of mammalian cells exposed to low-wavelength UVC light or simulated solar light [≤10% UVB (≥295 nm) and ≥90% UVA light] has confirmed that experimental UV exposure primarily causes UV signature C > T (and CC > TT) mutations in dipyrimidine sequences (Olivier et al. 2014; Nik-Zainal et al. 2015; Kucab et al. 2019). However, the sensitivity of these studies to detect less frequent or rare classes of UV-induced mutations was limited by both the high frequency of background mutations in mammalian cells grown in cell culture (e.g. Kucab et al. 2019) and the fact that that these studies typically only subjected cells to a single low-dose UV exposure prior to genome sequencing. Noncanonical substitutions in individual reporter genes have been detected in UV-irradiated cells (e.g. Armstrong and Kunz 1990; Sage et al. 1996; Kozmin et al. 2003), although due to the small numbers of these mutations studied, the potential mutagenic mechanism(s) involved were in many cases unclear.

Budding yeast grown in culture have much lower rates of background mutations and are more resistant to high doses of UV exposure than mammalian cells. Using budding yeast as a model system, we have recently observed that repeated exposure to UVC light (∼254 nm wavelength) induces not only canonical UV signature mutations but also noncanonical substitutions, including abundant T > C and T > A substitutions that resemble a number of driver mutations in melanoma (Laughery et al. 2020). These noncanonical mutation classes are elevated on the nontranscribed strand (NTS) of yeast genes (i.e. show transcriptional asymmetry; Pleasance et al. 2010; Haradhvala et al. 2016) and are significantly increased in nucleotide excision repair (NER)–defective cells, indicating that they are likely caused by helix-distorting UV photoproducts (Laughery et al. 2020). While these findings indicate that UVC-induced photoproducts cause noncanonical mutation classes, the extent to which these are induced by incident solar radiation, which is comprised solely of UVB (∼5–10% of solar UV) and UVA (∼90–95%) light, is unclear. UVB light predominantly induces CPDs and 6-4PPs that cause UV signature mutations (Pfeifer et al. 2005; Ikehata and Ono 2011; Brash 2015), while the exact mutation spectrum caused by UVA is controversial. Some studies have reported that UVA can induce low levels of CPDs and 6-4PPs and the resultant UV signature mutations (e.g. Rochette et al. 2003; Moreno et al. 2020), while others have indicated that UVA often (or even primarily) causes C > A/G > T substitutions, which may be a by-product of oxidative DNA lesions such as 8-oxoguanine (8-oxoG) (Kozmin et al. 2005; Yurchenko et al. 2022). To what extent UVA- or UVB-induced photoproducts cause noncanonical T > A or T > C substitutions is unclear.

Helix-distorting UV photoproducts (e.g. CPDs and 6-4PPs) are removed by NER. NER consists of 2 subpathways: the global genomic NER (GG-NER), which repairs helix-distorting DNA lesions throughout the genome, and transcription-coupled NER (TC-NER), which rapidly removes damage from the transcribed strands (TS) of genes (Scharer 2013; Lans et al. 2019). Consequently, there are typically fewer mutations derived from UV photoproducts on the TS of genes relative to the NTS, which is termed transcriptional asymmetry (Pleasance et al. 2010; Zheng et al. 2014; Haradhvala et al. 2016). Certain classes of mutations are also enriched on the leading or lagging strand during replication, in what is known as replication asymmetry (Haradhvala et al. 2016; Hoopes et al. 2016; Seplyarskiy et al. 2016; Tomkova et al. 2018). While replication asymmetry is often associated with mutations caused by APOBEC cytidine deaminases, which preferentially target single-stranded DNA regions during lagging strand replication (Haradhvala et al. 2016; Hoopes et al. 2016; Seplyarskiy et al. 2016), it may also impact oxidative mutations and other mutation classes, such as those induced by UV exposure (Tomkova et al. 2018).

Here, we characterize the spectrum of mutations associated with solar UV exposure by sequencing the genomes of yeast isolates repeatedly irradiated with either UVA or UVB light. Analysis of ∼20,000 mutations derived from wild-type (WT) or repair-deficient yeast cells indicates that UVB induces not only canonical UV signature mutations but also noncanonical C > A, T > A, and T > C substitutions. Our data further indicate that UVA-irradiated yeast cells, particularly in base excision repair (BER)-defective ogg1Δ mutants, display a unique spectrum of G > T substitutions, which unlike UVC- and UVB-induced mutations are specifically elevated on the leading strand during DNA replication. These findings reveal distinct mutagenic pathways associated with UVA and UVB light and suggest a possible mechanism by which solar radiation may induce key oncogenic mutations in melanoma.

Materials and methods

Yeast strains

Diploid yeast (Saccharomyces cerevisiae) strains were used in all passaging experiments to promote survival and allow for the unbiased accumulation of UV-induced mutations (i.e. in essential genes). Haploid yeast strains were used for all other assays. Strains were constructed by homologous recombination-mediated replacement of selective markers (e.g. TRP1, URA3, and KanMX) for targeted genes in BY4741 and BY4742 backgrounds (Brachmann et al. 1998), which were then mated to produce the resultant diploid strains. Diploids were screened by plating on selective media and mutants were also checked by PCR analysis. WT and rad16Δ strains have been previously described (Laughery et al. 2020), and ogg1Δ was constructed by replacement of OGG1 with the LEU2 and URA3 genes in BY4741 and BY4742, respectively.

Haploid strains were utilized for UV sensitivity, canavanine resistance, and C > A reporter assays. The rad16Δ mutant strain has been previously described (Laughery et al. 2020). The ogg1Δ haploid strain was constructed by targeted gene replacement of OGG1 with LEU2 in BY4741 yeast. The pcd1Δ and ntg1Δntg2Δ strains were constructed in the MP019 parental strain by replacement of the targeted gene (i.e. PCD1 and NTG1, respectively) with TRP1; NTG2 was replaced with LEU2 in the double knockout.

The dsDNA C > A reporter strain was constructed in ySR128 (Roberts et al. 2012). First, an ogg1Δ mutant was created by targeted insertion of KanMX into the gene locus and selection on G418 medium. Next, the ura3 K93Q mutation was introduced in the ogg1Δ strain via transformation of the pML104-based Cas9/CRISPR plasmid, pML156, and hybridized OML338/339 template oligonucleotides and confirmed by plating on media containing 5-fluoroorotic acid (5-FOA) and sequencing analysis (Laughery et al. 2015). The WT ura3 K93Q strain was constructed by mating ogg1Δ K93Q yeast with yBV228 or yBV229 and selecting for Cdc13-1+ and G418+ spores. Genotypes were further confirmed by plating on synthetic complete plates lacking uracil (SC-Ura) and PCR analysis.

Single-stranded DNA reporter strains were generated by transformation of a purified PCR product into cdc13-1 mutant strains yBV039 and yDM14, which contain the URA3 gene relocated to the subtelomeric region of chromosome V in forward and reverse orientations, respectively. Genomic DNA from the ogg1Δ ura3 K93Q reporter strain was used as the template for amplification with oligos OML270 and OML271. Transformed cells were plated on 5-FOA media to select for ura3Δ colonies, and proper editing was confirmed via sequencing analysis of the mutation locus.

UV irradiation of yeast

All UVB irradiations throughout this study were carried out in a UVP CL-1000M midrange crosslinker (Analytik Jena), which contains bulbs with peak UV emission at 302 nm. The lamp emission spectrum has negligible intensity below 280 nm. To perform irradiations, the crosslinker was programmed to the desired dosage and plates were irradiated with their lids off. For the WT passaged dose of 300 J/m², the UV exposure lasted approximately 11 s.

All UVA irradiations were performed in a UVP CL-1000 L longwave crosslinker (Analytik Jena), which contains bulbs with peak UV emission at 365 nm. Exposure time for UVA radiation was determined by timing a lower preset dose on the crosslinker and calculating the corresponding time required for 100, 200, and 300 kJ/m² doses. All exposures were performed with lids on petri dishes, which prevents exposure to any residual short-wavelength UV light (i.e. UVB or UVC). The passaged dose of 200 kJ/m² lasted ∼87 min. For reversion assays, plates were exposed while on ice.

UV sensitivity assays

Yeast strains were cultured in yeast peptone dextrose (YPD) medium and then diluted in water before plating on YPD plates. Once dry, plates were exposed to 150, 300, 450, or 600 J/m² UVB. Plates were then incubated at 30°C and colonies were counted as they became visible over a course of 8–11 days, with the majority of colonies typically being present by days 5 and 6. Exposures and initial incubations were performed in a yellow light room to inhibit repair by endogenous photolyases. Plates were counted in a normally lit room starting at ∼2 days postexposure. Sensitivity was calculated as the number of colonies growing on UV-exposed plates divided by the colonies growing on unexposed plates, adjusted for the dilution factor of each. A minimum of 3 independently exposed replicates were performed for each dose.

UVA spotting assay

Yeast cells were grown overnight in YPD media, subcultured, and allowed to grow to midlog phase (OD₆₀₀ 0.6–0.8). The cells were then resuspended in YPD to an approximate concentration of 1 × 10⁸ cells/mL, after which a series of 5 10-fold dilutions was performed to obtain cultures with ∼10⁷, 10⁶, 10⁵, 10⁴, and 10³ cells/mL. Five microliters of each dilution (10⁸ … 10³ cells/mL) was spotted sequentially on a YPD plate and briefly allowed to dry before irradiation with the indicated dose of UVA. The YPD plates were then incubated at 30°C and allowed to grow for 2–3 days before imaging.

UVB passaging assays

Yeast strains grown to late log phase were harvested by centrifugation, resuspended in water to ∼1 × 10⁷ cells/mL, and spotted on YPD plates. Once dry, plates were exposed to 150 J/m² UVB (rad16Δ) or 300 J/m² (all other strains). Plates were then incubated at 30°C overnight or longer, minimally until growth appeared. Cells from each spot were individually resuspended in sterile water, spotted again on fresh YPD plates, and reexposed to the same dose of UVB. This process was repeated for a total of 15 UV irradiations. Following this process, all spots were struck for isolation to fresh YPD plates and individual isolates from each spot were patched to new plates. Genomic DNA was extracted and submitted to whole genome sequencing as previously described (Laughery et al. 2020). A second set of WT yeast was repeatedly exposed to the same dose of UVB, but was replica plated to a new plate before each UVB exposure and the genomes of individual isolates were sequenced.

UVA passaging assays

Yeast cells were grown overnight in YPD media, subcultured, and allowed to grow to midlog phase (OD₆₀₀ 0.6–0.8). Yeast cells were then resuspended to an approximate concentration of 1 × 10⁷ yeast cells/mL in sterile water, spotted on YPD plates in a 5 × 5 grid (25 spots/plate), and allowed to dry. Plates were then irradiated with ∼200 kJ/m² UVA. The UV-treated plate was then incubated at 30°C overnight. Approximately 1 × 10⁷ yeast cells were taken from each spot, resuspended separately in sterile water, and respotted on fresh YPD plates prior to UVA irradiation as before. This process was repeated for a total of 15 UVA passages. Following the final UVA exposure and passaging, each yeast spot was struck for isolation onto YPD plates. Isolated single colonies from each sample were then patched to a fresh YPD and subsequently used for genomic DNA isolation followed by whole genome sequencing as previously described (Laughery et al. 2020).

Canavanine resistance assays

Yeast strains were cultured in YPD medium, harvested by centrifugation, and resuspended in sterile water or phosphate-buffered saline (PBS) to an approximate density of 1 × 10⁶ cells/mL. Cell suspensions were then poured into small petri dishes and irradiated with ∼100 kJ/m², 200 kJ/m², or 300 kJ/m² UVA (for UVA experiments, doses based on the timing required according to the manufacturer's specifications) or 150 J/m² or 300 J/m² (for UVB experiments). Control cells were not UV irradiated. After exposure, cells were collected by centrifugation, resuspended in YPD medium, and incubated in a 30 °C shaker for >6 h to allow for resumption of cellular replication. UVB exposures and outgrowths were performed in a yellow light room to avoid activation of endogenous yeast photolyases, and UVA exposures were performed in normal room lighting conditions. Following outgrowth, cells were diluted in sterile water or PBS and spread onto SC-Arg + 0.006% (w/v) canavanine plates to assess mutation frequency of the CAN1 reporter gene. Cell dilutions were also plated SC media to estimate the total cell number. Mutation frequency was calculated as the proportion of colonies growing on SC-Arg + Can plates relative to the total colony number on control plates, each adjusted for their dilution factor. A minimum of 6 independently UV-exposed replicates were performed for each strain.

dsDNA reporter assays

Reporter strains were cultured overnight in YPD or yeast extract peptone dextrose adenine (YPDA) media before harvesting cells by centrifugation, resuspension in water, and spreading onto SC-Ura. Dilutions of each cell suspension were also made in water and spread onto SC plates to assess the approximate total number of viable cells plated to calculate reversion frequencies. Plates were exposed to ∼200 kJ/m² of UVA on ice (based on timing according to the manufacturer's specifications) and then incubated at 30°C for 3 days before counting colonies on both selective and control plates. A minimum of 6 replicates were performed for each strain. The reversion frequency was calculated as the number of colonies on SC-Ura plates divided by the number of colonies on SC plates multiplied by their dilution factor.

To confirm the mutation status of yeast growing on SC-Ura following UVA exposure, isolated colonies were patched to fresh SC-Ura plates, incubated to allow growth, and genomic DNA was extracted by bead beating and phenol:chloroform:isoamyl alcohol extraction followed by ethanol precipitation. Precipitated DNA was resuspended in Tris-EDTA buffer (pH 7.5), RNAse A treated, and the URA3 K93 mutation locus was PCR amplified using EconoTaq DNA Polymerase (Lucigen) with the OML270 (GCCGCTAAAGGCATTATCCG) and OML271 (CAATGCGTCTCCCTTGTCATC) primers. PCR products were column purified (Zymo Research) and submitted for Sanger sequencing. Eleven isolated revertants were sequenced to confirm the C > A mutations.

ssDNA reporter assays

Yeast strains were cultured overnight in YPDA medium at 23°C. Cells were then diluted 6-fold in 37°C prewarmed YPDA media and incubated at 37°C for 6 h to induce G2/M arrest, evidenced by dumbbell-shaped cell morphology when observed under a microscope. Cells were then collected by centrifugation, washed with sterile water, and resuspended in water. Cell suspensions were exposed to ∼200 kJ/m² UVA in petri dishes on ice (see above) before harvesting by centrifugation and spreading onto SC-Ura plates to screen for revertant colonies. Dilutions were also made and plated on SC medium to calculate the total number of cells plated on selective medium. Plates were incubated at ∼23°C for 5 days (or 4–5 days for SC plates) prior to colony counts. Reversion frequencies were calculated as the number of colonies on SC-Ura plates divided by the number of colonies on SC plates multiplied by their dilution factor. A minimum of 6 replicates were performed for each strain.

Bioinformatics analysis

Mutations were called from whole genome sequencing data derived from each yeast isolate using CLC Workbench version 7.5, as previously described (Mao et al. 2017; Laughery et al. 2020), with the exception that only unique mutations across all UVB-irradiated isolates (or across all UVA-irradiated and no-UVA isolates) were retained (instead of ≤4 isolates containing the mutation for UVC data; including mutations in multiple isolates only increased the mutation count by ∼3%). Mutations mapping to the mitochondrial chromosome were excluded and 1 isolate with very low sequencing depth was not analyzed further. Bioinformatics analysis of genome-wide maps of UVB- or UVA-induced mutations, including analysis of transcriptional asymmetry and mutations per isolate, was performed using custom Perl scripts and the Saccer3 genome sequence, as previously described (Laughery et al. 2020). Significant changes in mutation frequency for individual mutation classes in UVB- or UVA-exposed cells relative to the no UV control were determined by t test with either Bonferroni (UVB) or Holm–Sidak (UVA) correction for multiple hypothesis testing. The frequency of mutations on the TS and NTS of yeast genes was normalized by frequency of each trinucleotide sequence context in TS and NTS of genes throughout the yeast genome. Gene coordinates were derived from Park et al. (2014). Replication asymmetry was characterized by 2 methods. First, mutations within 8 kb of a replication origin, using a published map of 317 replication origins (Nieduszynski et al. 2006; Soriano et al. 2014), were assigned to the leading or lagging strand template based on the expected direction of replication across these 8-kb regions and the bases on which DNA lesions were expected to form. Mutation frequency on the leading and lagging strand was normalized based on the frequency of each sequence context on the leading and lagging strands within 8 kb of a replication origin. Significance of replication asymmetry was determined using chi square test with a Bonferroni correction for multiple hypothesis testing. Second, the positions of confirmed yeast replication origins listed in OriDB (Nieduszynski et al. 2007) (http://cerevisiae.oridb.org/index.php) were used to determine factional interorigin zones. Regions between each neighboring origin were subdivided into zones corresponding to one-tenth the fractional distance. Mutations were assigned to each zone and the number of C to A or G to T substitutions per zone counted. Corresponding substitution types for similar fractional zones were aggregated and the percent of complementary substitutions calculated as in Hoopes et al. (2016).

Results

To characterize the mutation spectra resulting from exposure to UV wavelengths found in solar radiation, we passaged diploid yeast (S. cerevisiae) through 15 rounds of exposure to either UVB or UVA light (followed by cell regrowth/recovery) and then sequenced the genome of individual clonal isolates to identify the resulting mutations (Fig. 1a). We chose a dose of 300 J/m² of UVB light since this did not drastically affect cell survival (Fig. 1b) and caused a similar induction of CAN1 mutation frequency (Supplementary Fig. 1a and b) to the UVC dose (25 J/m²) used in our previous study (Laughery et al. 2020). For UVA irradiation, we chose a dose of 200,000 J/m², since higher doses (e.g. 300 kJ/m²) resulted in significant cell killing (Supplementary Fig. 2). This UVA dose elevated the frequency of canavanine-resistant (Can^R) yeast (Supplementary Fig. 1c), consistent with a previous report (Kozmin et al. 2005), although mutation induction by UVA was significantly lower (>10-fold less) than for the UVB or UVC doses used (Supplementary Fig. 1a–c). Genome sequencing of independent isolates revealed an induction of mutations following repeated UVB exposure across the different isolates, roughly equivalent to the mutation frequencies observed following UVC treatment (Fig. 1c). Comparison with the untreated (“no UV”) control isolates (Fig. 1c) indicates that ∼99% of the mutations in the UVB isolates are UV-induced. In contrast, UVA exposure caused a modest, albeit significant (P < 0.05), increase in mutations relative to the “no UV” control (Fig. 1c). Comparison of the number of mutations induced across the yeast genome per kJ/m² of each type of UV wavelength indicates that UVB light is ∼10-fold less mutagenic than UVC, while UVA light is >100,000-fold less mutagenic than UVC in yeast (Fig. 1d). This 10-fold difference in UVB and UVC mutagenicity is consistent with previous reports showing that UVC induces CPDs and 6-4PPs ≤10-fold more frequently than UVB (Mitchell et al. 1991, 1992; Douki and Cadet 2001).

Fig. 1.

a) Diagram of repeated UV irradiation passaging procedure. Yeast strains were spotted on plates, exposed to UV light, and allowed to regrow. After 15 repetitions of this procedure, whole genome sequencing was performed on individual clonal isolates. Image adapted from Laughery et al. (2020). b) Quantitative UV survival assay of haploid WT yeast cells following the indicated doses of UVB irradiation; 300 J/m² of UVB light was the chosen dose for passaging WT cells. c) Mutation frequency per isolate of WT yeast cells following repeated exposure (15×) of indicated doses of UVA, UVB, or UVC light. Only WT UVB isolates from the first passaging assay are shown. UVC mutation data from Laughery et al. (2020). Zero mutation values indicate the absence of any mutations in these isolates. d) Mutation frequency in yeast per kJ/m² of UV used for irradiation.

UVB light induces noncanonical mutation classes

Analysis of the trinucleotide mutation spectra of the aggregated UVB-exposed isolates (Fig. 2a) revealed not only abundant canonical UV mutations (i.e. C > T substitutions in dipyrimidine sequences), but also frequent noncanonical mutation classes. These include C > A, T > A, and T > C substitutions, which comprise 4.9%, 14%, and 36%, respectively, of the 6,530 total single nucleotide substitutions in UVB-exposed isolates. This UVB mutation spectrum was highly reproducible, based on an internal comparison with the mutation spectrum of independent UVB exposure experiments (Supplementary Fig. 3), and closely resembles our previously published UVC mutation spectrum in yeast (cosine similarity = 0.98). Since UV-induced CPDs and 6-4PPs form exclusively at dipyrimidine sequences, we examined the number of mutations per isolate for each of these substitutions (e.g. C > A, C > T, T > A, and T > C) occurring in either the 5′ or 3′ position of a dipyrimidine or in a nondipyrimidine sequence. We compared the number of mutations per isolate for each of these mutation classes in the UVB isolates to the no UV control isolates using a stringent cutoff (see Materials and methods) to determine which mutation classes are significantly elevated by UV exposure.

Fig. 2.

UVB exposure induces both canonical and noncanonical UV mutation classes. a) Trinucleotide contexts of aggregated single nucleotide mutation spectra for WT yeast cells following the UV passaging protocol using 300 J/m² of UVB. Mutations from both WT UVB replicate passaging experiments are depicted. b) Mutation frequency per isolate classified by substitution class and by position of mutation within dipyrimidine context for UVB-treated WT isolates relative to the no UV control. Whether the mutation class matches a canonical UV mutation class (i.e. C > T in dipyrimidine sequence) or noncanonical UV mutation class is indicated. Mean ± SEM is depicted. **P < 0.00001 and *P < 0.001, based on t test using Bonferroni correction for multiple hypothesis testing. c) Transcription asymmetry analysis of UVB-induced mutations in WT yeast cells. Normalized fraction of mutations on the TS and NTS of yeast genes is depicted. *P < 0.05 and **P < 0.001 based on chi square test with Bonferroni correction. d, e) Plot of transcription asymmetry (normalized by frequency of each trinucleotide sequence context in TS and NTS of yeast genes) for UVB-induced mutations in WT yeast cells (both replicates) plotted for individual trinucleotide sequence contexts, plotted relative to the total number of mutations of each trinucleotide class. Only mutation classes with at least 29 mutations were plotted. Mutations occurring in a dipyrimidine sequence context are indicated with a circle with a solid outline. The color of the circle indicates the mutation type (e.g. C > T).

The results indicate that C > T mutations are significantly elevated (P < 0.001) at both the 5′ and 3′ positions of dipyrimidine sequences (Dipyr) in UVB-exposed cells relative to the no UV control (Fig. 2b), while C > T mutations at nondipyrimidine sequence contexts are not significantly elevated, as expected. Notably, the same pattern is apparent for C > A substitutions, which are significantly elevated in dipyrimidine contexts, but not in nondipyrimidine sequence contexts. Analysis of the mutation frequencies of C > A or C > T substitutions in different sequence contexts indicates that these mutation classes are highly correlated in UVB-exposed cells (cosine similarity = 0.865; Pearson correlation coefficient = 0.745; P = 0.001), suggesting that they might represent different mutational outcomes (i.e. C > A or C > T) arising from replicative bypass of the same UV lesion(s). In contrast, C > G mutations are not significantly elevated in UVB-exposed cells relative to the no UV control (Fig. 2b). T > A, T > C, and T > G mutations are also elevated (P < 0.001) in UVB-exposed isolates relative to no UV isolates, except these mutation classes were significantly elevated in both dipyrimidine and nondipyrimidine sequence contexts (Fig. 2b). UVB-induced T > C substitutions are most enriched at the 3′ position of dipyrimidines, while T > A and T > G mutations occurred with similar frequency at 5′ or 3′ dipyrimidines or nondipyrimidine contexts (Fig. 2b), suggesting that these latter mutation classes are not primarily caused by lesions occurring at dipyrimidines. Overall, the UVB mutation spectra closely resemble that of UVC, although closer inspection revealed subtle differences, with significantly more C > T substitutions and fewer T > C and T > A substitutions in UVB-exposed yeast (Supplementary Fig. 4).

Mutations arising from helix-distorting UV photoproducts typically occur less frequently on the TS than the NTS due to rapid repair by the TC-NER pathway. We tested whether each mutation class (e.g. C > A, C > T, and T > A) showed the transcriptional asymmetry expected for UV photoproducts by determining the ratio of mutations on the NTS to TS across all yeast genes. This analysis indicated that C > T substitutions in the UVB-exposed yeast are significantly elevated (∼2.2-fold, P < 0.001) on the NTS relative to the TS of yeast genes (Fig. 2c), consistent with the hypothesis that these C > T mutations primarily arise from helix-distorting CPDs or 6-4PPs. C > A mutations are also significantly elevated on the NTS (Fig. 2c; P < 0.05), suggesting that many of these noncanonical mutations also arise from helix-distorting UV photoproducts. C > G mutations also showed transcriptional asymmetry, but there were too few of this class of mutations for this difference to be significant. Analysis of transcriptional asymmetry for these mutation classes in individual trinucleotide contexts (with at least 29 total mutations) indicated that the most frequent C > A mutation classes occur in dipyrimidine sequences (e.g. TCN) and are elevated on the NTS of yeast genes (with the exception of the CCA context), roughly similar to the transcriptional asymmetry of C > T mutations (Fig. 2d).

Similar analysis of UVB-induced mutations at thymine bases revealed that T > C substitutions are significantly elevated on the NTS (∼2.4-fold; P < 0.001), while T > A and T > G mutations in aggregate do not show transcriptional asymmetry (Fig. 2c). Analysis of the transcriptional asymmetry of individual trinucleotide contexts confirmed that T > C mutations in dipyrimidine contexts displayed consistently more mutations on the NTS, consistent with these mutations arising from helix-distorting UV photoproducts (Fig. 2e), such as 6-4PPs (Bresson and Fuchs 2002). In contrast, T > A substitutions showed strikingly different transcriptional asymmetry, depending on the specific sequence context (Fig. 2e). T > A substitutions in TTC, TTG, or TTT contexts are elevated on the NTS, similar to T > C substitutions; in contrast, T > A substitutions in ATA, CTA, GTA, and TTA contexts occur less frequently on NTS relative to the TS (Fig. 2e). These opposite patterns of transcription asymmetry in different sequence contexts can explain why T > A mutations in aggregate lack transcriptional asymmetry. Closer inspection of Fig. 2e indicates that only T > A mutations occurring in an NTA motif showed reverse transcriptional asymmetry, suggesting that the actual mutation in these contexts is an A > T substitution in a TAN motif occurring on the opposite DNA strand (i.e. the NTS; Fig. 2e). Taken together, these findings indicate that T > A substitutions in a TTN context (with the exception of TTA) are likely caused by mutagenic bypass of a canonical UV photoproduct (i.e. CPD or 6-4PP), while T > A substitutions in an NTA sequence context are actually an A > T substitution on the opposite strand, likely caused by an atypical photoproduct occurring in TAN sequences.

UVB mutation spectra in GG-NER–deficient yeast cells

If the noncanonical mutation classes (e.g. C > A, T > C, and T > A substitutions) induced by UVB exposure are caused by helix-distorting UV photoproducts, they should also be elevated in NER-deficient cells. To test this hypothesis, we exposed yeast cells deficient in the GG-NER pathway (i.e. rad16Δ mutant cells, see Fig. 3a) to 15 doses of UVB light and identified UVB-induced mutations by genome sequencing of independent isolates. Due to the elevated UV sensitivity of rad16Δ yeast (Fig. 3b), we used a lower UVB dose of 150 J/m². This dose still resulted in an elevated mutation frequency relative to the WT control across the genome of individual yeast isolates (Fig. 3c), consistent with our CAN1 mutation data (Supplementary Fig. 5). To determine which UVB-induced mutation classes are elevated in the NER-deficient cells, we compared the number of mutations per isolate from each mutation class in rad16Δ relative to WT (Fig. 3d). Here, we observed a ∼2–3-fold higher frequency of canonical C > T mutations in a dipyrimidine context in the rad16Δ mutant relative to WT (Fig. 3d), consistent with these mutations arising from canonical UV photoproducts that are repaired by GG-NER. However, noncanonical C > A, C > G, T > A, T > C, and T > G substitutions are also significantly elevated in dipyrimidine contexts (P < 0.05) in the rad16Δ mutant (Fig. 3d), suggesting that these noncanonical substitutions may be caused by mutagenic bypass of UV-induced CPDs or 6-4PP. Notably, T > A substitutions in a nondipyrimidine context are also significantly elevated (P < 0.05; Fig. 3d), suggesting that these mutations may be caused by an atypical UV photoproduct.

Fig. 3.

a) Diagram illustrating the key role of Rad16 in promoting repair via the GG-NER pathway on the NTS of genes and within intergenic regions. The TS is primarily repaired by the TC-NER pathway. b) Quantitative UV survival assay of WT and rad16Δ yeast cells following treatment with indicated doses of UVB light; 150 J/m² was the chosen dose for rad16Δ passaging due to the UV sensitivity of this repair-deficient strain. c) Mutation frequency per isolate of WT and rad16Δ yeast strains following UVB passaging. d) Mutation frequency per isolate classified by substitution class and by position of mutation within dipyrimidine context for UVB-treated WT and rad16Δ yeast cells. Mean ± SEM is depicted. **P < 0.0001 and *P < 0.01, based on t test using Holm–Sidak correction for multiple hypothesis testing. e) Transcription asymmetry analysis of UVB-induced mutations in rad16Δ yeast cells. Normalized fraction of mutations on the TS and NTS of yeast genes is depicted. **P < 0.001 based on chi square test with Bonferroni correction. f, g) Transcription asymmetry analysis (normalized by frequency of each trinucleotide sequence context in TS and NTS of yeast genes) of individual f) C > A, C > T, and g) T > A mutation classes, separated by trinucleotide context, plotted relative to the total number of mutations of each mutation class in rad16Δ yeast cells. Mutations occurring in a dipyrimidine sequence context are indicated with circle with a black outline. h) Same as e) with only the normalized fraction of T > A mutations (distinguished by trinucleotide context) plotted on the TS and NTS of yeast genes. T > A substitutions occurring in a NTA sequence context are enriched on the TS of yeast genes, indicating that the causative lesion occurs on the opposite strand (NTS) and therefore is an A > T substitution in a TAN sequence context. *P < 0.05 and **P < 0.001 based on chi square test with Bonferroni correction. i) Frequency per isolate of A > T substitutions in a TAN sequence context for WT and rad16Δ yeast strains following UVB and UVC passaging. Mean ± SEM is depicted. **P < 0.001, based on t test using Holm–Sidak correction for multiple hypothesis testing.

Analysis of transcriptional asymmetry of C > T substitutions in aggregate indicates that the elevated mutation frequency on the NTS (relative to the TS) of yeast genes in WT cells (e.g. Fig. 2c) is more pronounced in rad16Δ cells (Fig. 3e), presumably because the rad16Δ mutant is unable to repair lesions on the NTS (Fig. 3a). A nearly identical trend was observed for T > C substitutions, confirming that this mutation class is likely caused by a canonical UV photoproduct (i.e. CPDs or 6-4PPs). Less abundant mutation classes such as C > A, C > G, and T > G substitutions are also significantly elevated in the rad16Δ cells and are significantly enriched on the NTS (Fig. 3e). The most abundant of these are C > A substitutions in TCN sequence context, which are highly enriched on the NTS of yeast genes relative to the TS in rad16Δ cells, similar to C > T substitutions (Fig. 3f). Since these mutations are associated with lesion-forming dipyrimidine sequences, these findings indicate that mutagenic bypass of a CPD or 6-4PP may occasionally result in C > A substitutions.

The most abundant T > G substitution class occurs in a GTT sequence context. This mutation class is highly enriched on the NTS of yeast genes (Supplementary Fig. 6a), similar to UV-induced C > T mutations. T > A and T > C mutations also occur in this sequence context in UVB-exposed WT or rad16Δ cells, but these are less abundant than T > G mutations (Supplementary Fig. 6b and c). In contrast, T > A mutations are most abundant in an ATT sequence context and T > C mutations are most abundant in a CTT sequence context (Supplementary Fig. 6b and c). These findings suggest that mutagenic bypass of the 5′ base of a presumptive TT photoproduct favors misinsertions that could pair with the flanking base, perhaps via a template slippage model.

UV-induced T > A substitutions showed a different pattern of transcriptional asymmetry. Whereas T > A mutations in aggregate did not exhibit strand asymmetry in WT cells (Fig. 2c), they occur more frequently on the TS in rad16Δ mutant cells (Fig. 3e). Notably, this is the opposite pattern observed for other mutation classes. Analysis of individual T > A mutation classes revealed that T > A substitutions in TTT, TTG, and TTC trinucleotide contexts displayed normal transcriptional asymmetry in rad16Δ cells, while T > A substitutions in NTA sequence context show the opposite pattern, with elevated mutation frequencies on the TS relative to the NTS (Fig. 3g and h). This closely matches the pattern observed in WT cells, although the degree of transcriptional asymmetry is exaggerated in the absence of GG-NER. These findings are consistent with the model that the causative mutagenic lesion actually occurs on the opposite strand at TAN sequences on the NTS and causes A > T substitutions (see lower panel of Fig. 3h), which matches the mutation signature of an atypical TA photoproduct (Bose et al. 1983; Zhao et al. 1996; Zhao and Taylor 1996; Laughery et al. 2020). Further analysis indicates that UVB induces abundant A > T substitutions that fit this mutation signature and that this mutation class is elevated in repair-deficient rad16Δ cells (Fig. 3i). These findings suggest that UVB can induce mutations that match a mutation signature associated with TA photoproducts, albeit somewhat less efficiently than UVC light.

UVA exposure causes a distinct mutation signature, particularly in cells lacking the Ogg1 DNA glycosylase

Exposure to repeated doses of UVA light (i.e. 15 × 200 kJ/m²) resulted in a slight, but significant (P < 0.05), increase in mutation rates across the yeast genome (Fig. 1c). Analysis of the resulting mutation spectra indicates that UVA exposure of WT cells induces C > A and C > G substitutions relative to unexposed control cells (Fig. 4a). This mutation spectrum is distinct from that of UVB and UVC exposure, but is consistent with prior findings that UVA exposure primarily induces oxidative lesions in yeast (Kozmin et al. 2005). To test this hypothesis, we analyzed the frequency of UVA-induced mutations in a CAN1 reporter assay in yeast strains deficient in repair pathways that remove oxidative DNA damage. Notably, deletion of OGG1, which encodes a glycosylase that functions in the repair of the oxidative lesion 8-oxoG (Friedberg et al. 2006), significantly elevated the frequency of UVA-induced Can^R mutants (Fig. 4b), in accordance with a previous report (Kozmin et al. 2005). In contrast, deletion of NTG1 and NTG2, which repair oxidized pyrimidine bases (Sentürker et al. 1998; Alseth et al. 1999; Friedberg et al. 2006), or PCD1, which is purported to function in sanitizing the nucleotide pool of oxidized purines (Nunoshiba et al. 2004), did not significantly affect the frequency of UVA mutation frequency (Fig. 4b). These findings indicate that UVA exposure induces mutagenic 8-oxoG lesions in DNA, which are normally suppressed in WT cells by the Ogg1 repair protein.

Fig. 4.

a) Single nucleotide substitution spectra for WT yeast cells following passaging with or without exposure to UVA. **P < 0.001, based on t test using Holm–Sidak correction for multiple hypothesis testing. b) CAN1 mutation frequency as measured by canavanine resistance (Can^R) of WT, ntg1Δntg2Δ, ogg1Δ, or pcd1Δ yeast strains following UVA irradiation. c) Mutation frequency per isolate in WT and ogg1Δ yeast strains following passaging with or without UVA treatment. Zero mutation values indicate the absence of any mutations in these isolates. d) Mutation frequency per isolate classified by substitution class for UVA-treated and untreated ogg1Δ yeast. **P < 0.001, based on t test using Holm–Sidak correction for multiple hypothesis testing. e) Number of C > A single nucleotide substitutions in UVA-treated ogg1Δ yeast cells classified by trinucleotide sequence context. C > A mutations are enriched at ACA, ACN, and NCA sequences. f) Graphical representation of overrepresented and underrepresented sequences flanking G > T (i.e. C > A) substitutions in UVA-treated ogg1Δ yeast. Data plotted as log odds of binomial probability of each nucleotide. Panel generated using pLogo (O'shea et al. 2013).

To analyze the genome-wide distribution of UVA-induced mutations in repair-deficient cells, we exposed ogg1Δ mutant cells to 15 doses of 200 kJ/m² of UVA light. Similar to WT cells, this dose of UVA did not cause significantly killing in ogg1Δ cells (Supplementary Fig. 2). Genome sequencing of individual isolates revealed a consistent induction of mutations following UVA exposure in ogg1Δ cells that is elevated ∼10-fold relative to both UVA-exposed WT cells and unexposed ogg1Δ cells (Fig. 4c). Nearly all of the mutations in UVA-exposed ogg1Δ cells are C > A/G > T substitutions (Supplementary Fig. 7a), which are significantly elevated relative to the unexposed ogg1Δ cells (Fig. 4d). UVA exposure of ogg1Δ cells also increases the frequency of C > G substitutions (Fig. 4d, see inset), but these comprise only a minor fraction of UVA-induced mutations.

Analysis of the combined trinucleotide mutation spectra of UVA-exposed ogg1Δ isolates revealed that C > A mutations are most frequently associated with ACA sequences, occurring >3-fold more abundantly than any other trinucleotide context (Fig. 4e and Supplementary Fig. 7a). C > A mutations are also abundant in ACN or NCA sequence contexts (Fig. 4e). Since Ogg1 repairs 8-oxoG lesions (Friedberg et al. 2006), the causative lesion likely occurs on the opposite, guanine-containing strand and reflects a unique mutation signature in which G > T mutations primarily occur at sites with a flanking thymine base (i.e. TGN or NGT), whereas other flanking nucleotides (i.e. C, G, or A) are depleted (Fig. 4f). This mutation pattern is also apparent in ogg1Δ cells that were not exposed to UVA (Supplementary Fig. 7b–d), although the enrichment of flanking thymines is less prominent (Supplementary Fig. 7d). These results indicate that in repair-deficient cells, UVA primarily induces C > A/G > T substitutions in guanine bases flanked by on 1 or both sides by thymine.

UVA-induced mutations in ogg1Δ cells show replication asymmetry, not transcriptional asymmetry

Since UVB-induced mutations occur less frequently on the TS due to TC-NER, we tested whether UVA-induced mutations show similar transcriptional asymmetry. Analysis of the compiled C > A mutations in UVA-irradiated ogg1Δ cells indicated roughly equivalent numbers on both the TS and the NTS of yeast genes (Fig. 5a). In contrast, a similar analysis of C > A or C > T mutations in UVB-irradiated WT cells showed ∼2- to 3-fold more mutations on the NTS (Fig. 5b). Statistical analysis confirmed that significant transcriptional asymmetry did not occur for any specific C > A trinucleotide context (Fig. 5c) nor for these mutations in aggregate in UVA-irradiated ogg1Δ cells.

Fig. 5.

a, b) Plot of transcription asymmetry for different cytosine mutation classes (i.e. C > A, C > G, and C > T) in the UVA-treated ogg1Δ and UVB-treated WT yeast strains. Transcription asymmetry was calculated as the normalized ratio of mutations occurring on NTS relative to the TS. Each circle shows mutation data for an individual trinucleotide sequence context (e.g. ACA). Circles are color-coded for their mutation type (e.g. blue represents C > A mutations) and have black outline if the trinucleotide sequence context contains a dipyrimidine. Only data for the first (primary) WT UVB replicate is shown. c) Normalized fraction of C > A mutations occurring on the NTS relative to TS of all yeast genes in the UVA-treated ogg1Δ yeast strain within each trinucleotide context. d, e) Plot of replication asymmetry for different cytosine mutation classes (i.e. C > A, C > G, and C > T) in d) UVA-treated ogg1Δ and e) UVB-treated WT yeast strains. Replication asymmetry was calculated as the normalized fraction of mutations occurring on the leading versus the lagging strand. A published map of replication origins (Nieduszynski et al. 2006; Soriano et al. 2014) was used to assign mutations within 8 kb of an origin to either the leading or lagging strand of the replication fork. Only data for the first (primary) WT UVB replicate are shown. f) Elevated C > A mutations on the lagging strand of UVA-treated ogg1Δ yeast cells likely reflect elevated G > T mutations on the leading strand, since ogg1Δ repairs 8-oxoG lesions occurring on the guanine base. **P < 0.001 based on chi square test with Bonferroni correction. g) Genome-wide asymmetry of G > T and C > A mutations between yeast origins of replication (Nieduszynski et al. 2007). The percent of total G > T or complementary C > A substitutions was calculated among one-tenth fractional interorigin zones. Favoring of G > T substitutions 3′ of origins and C > A substitutions 5′ of origins is consistent with higher numbers of G > T substitutions associated with the leading strand template.

Certain mutation classes have been previously shown to exhibit replication asymmetry (Haradhvala et al. 2016), in which mutations are elevated on either the leading or lagging DNA strand. We wondered whether UVA-induced mutations might be elevated on 1 DNA strand during replication. We used a published map of replication origins (Nieduszynski et al. 2006; Soriano et al. 2014) to assign mutations within 8 kb of a mapped origin of replication to either the leading or lagging strand of the replication forks. Interestingly, we found that significantly fewer C > A mutations occurred on the leading strand than on the lagging strand (Fig. 5d). In contrast, analysis of C > A or C > T mutations in UVB-irradiated WT cells showed roughly equivalent numbers of mutations occurring on both strands of the replication fork (Fig. 5e). In aggregate, there were significantly more C > A mutations on the lagging strand than the leading strand in UVA-irradiated ogg1Δ cells. Given Ogg1's cellular function in repairing 8-oxoG lesions, these C > A/G > T mutations are likely derived from damage originating in guanine bases on the opposite DNA strand (i.e. the leading strand), corresponding to elevated G > T substitutions on the leading strand (Fig. 5f).

To confirm this finding, we analyzed UVA-induced C > A/G > T substitutions in ogg1Δ cells based on their relative distance to neighboring replication origins (Fig. 5g, top panel). This analysis revealed an enrichment of C > A mutations to the right of a replication origin and a corresponding enrichment of G > T mutations to the left of replication origin (Fig. 5g). These findings indicate that G > T mutations are enriched on the leading strand template, consistent with our previous analysis. In summary, our findings indicate that UVA-induced mutations in BER-deficient cells have replication asymmetry, but not transcriptional asymmetry.

UVA mutation signature can explain the recurrent NRAS Q61K driver mutation in melanoma

One of the most recurrent driver mutations in melanoma is NRAS Q61K, occurring in ∼7% of cases ( AACR Project GENIE Consortium 2017). NRAS Q61K is caused by a C > A/G > T substitution in an ACA/TGT sequence context. Since this mutation matches the UVA signature, we wondered whether this mutation may be caused by exposure to the UVA spectrum of sunlight. Consistent with this hypothesis, analysis of somatic mutation data from the COSMIC database (Forbes et al. 2017; Laughery et al. 2020) indicates that NRAS Q61K mutations are significantly enriched in skin cancers relative to other cancer types (Fig. 6a). As a further test, we developed a reversion reporter in which an essential lysine residue in the yeast URA3 gene was mutated to glutamine (ura3 K93Q), rendering the strain unable to grow on media lacking uracil (URA-). Importantly, a C > A/G > T reversion in a TGT sequence context (Fig. 6b), mimicking the NRAS Q61K mutation, results in Ura+ yeast cells. We used this reversion reporter to test whether this specific mutation class could be induced by UVA exposure.

Fig. 6.

a) Graph showing the enrichment in skin cancers relative to other cancer types for different mutations in the NRAS oncogene. Color of the circle indicates the mutation type and size of the circle indicates the abundance of the mutation in skin cancers. Notably, the NRAS Q61K driver mutation, which corresponds to a C > A substitution in an ACA sequence context, is enriched in skin cancers. Mutation data obtained from COSMIC (Forbes et al. 2017). Panel taken from Laughery et al. (2020). b) Schematic showing the design of a C > A reversion reporter in the URA3 gene, in which the K93 active site residue (Miller et al. 2001) is mutated to glutamine, resulting in an inactive enzyme that is unable to promote uracil biosynthesis. Here, a C > A substitution in an ACA sequence context results in Q-to-K reversion, thereby rendering cells able to synthesize uracil (Ura+). This substitution mimics the NRAS Q61K driver mutation. c) Reversion frequency for WT or ogg1Δ C > A reporter strains treated with or without UVA, determined by the number of Ura+ colonies. A few of the no UVA WT replicates had no Ura+ colonies; the reversion frequency for these replicates was calculated as if a single Ura+ colony was present. **P < 0.01 based on Mann–Whitney test. d) Diagram of C > A/G > T ura3 reversion reporter in a cdc13-1 strain, which was utilized to determine the lesion-forming base (i.e. C or G). The ura3-K93Q mutant was integrated in both orientations relative to the telomere end. Incubation of cells at 37°C for 6-h results in 5′ end resection, so that only a single DNA strand of the ura3 gene (i.e. either the C- or G-containing strand, depending upon the ura3 gene orientation) was retained prior to UVA irradiation. e) Ura+ reversion frequency of ura3-K93Q cdc13-1 cells with or without UVA irradiation. Orientation of ura3 reversion reporter is indicated. One of the no UVA ura3 reverse replicates had no Ura+ colonies; the reversion frequency for this replicate was calculated as if a single Ura+ colony was present. **P < 0.01 and *P < 0.05 based on Mann–Whitney test.

Our data indicate that the frequency of Ura+ reversion events is low in the absence of UVA light (reversion frequency of 2.0 × 10⁻⁸) but elevated 10-fold upon UVA treatment (Fig. 6c). Inactivation of the Ogg1 glycosylase resulted in a ∼240-fold induction of reversion events in the absence of UVA and a ∼7,000-fold induction upon exposure to UVA (Fig. 6c). To confirm our results, we sequenced the mutation locus of 11 Ura + UVA-irradiated ogg1Δ colonies and found that all had the expected C > A mutation resulting in a Q93K reversion event.

While the high frequency of Ura+ reversion events in ogg1Δ cells suggests that the Q93K reversion is actually a G > T mutation arising from UVA-induced 8-oxoG lesion on the opposite strand, we wanted to directly test whether the mutagenic lesion occurs on the G or C base. To do this, we repeated the ura3 reversion assays in a cdc13-1 mutant strain (Nugent et al. 1996). Growing the cdc13-1 mutant at 37°C causes telomere uncapping and 5′→3′ resection of the chromosome ends (Booth et al. 2001), so that adjacent genes become transiently single-stranded. We developed cdc13-1 strains in which the ura3 K93Q reversion reporter was integrated adjacent to yeast telomere V-L in both orientations (i.e. ura3 forward and reverse; see Fig. 6d). Growing these strains at 37°C for 6 h prior to UV radiation results in the removal of one of the DNA strands of the ura3 reporter, dependent upon its orientation, so that the resulting UV lesions are formed in ssDNA (Fig. 6d). Because many ssDNA lesions are not efficiently repaired by BER, damage occurring on the exposed strand should cause very high mutation rates, even though resection of the ura3 gene occurs in only a subset of cells.

Using this system, we exposed temperature shifted (i.e. ssDNA induced) cdc13-1 cells to UVA light and measured the frequency of Ura+ reversion events. In the reverse orientation, the C-containing strand of the ura3 K93Q reporter was retained (Fig. 6d). Here, the frequency of Ura+ reversion events was similar to that observed for the double-stranded ura3 reporter (compare Fig. 6c and e). In contrast, when the G-containing strand of the ura3 K93Q reporter was retained in the forward orientation, the frequency of Ura+ reversion events was ∼19-fold higher (P < 0.001) than the double-stranded reporter and also significantly enriched (∼7-fold) relative to unirradiated cells (P < 0.001; Fig. 6e). Taken together, these data indicate that UVA primarily induces G > T substitutions by creating lesions (likely 8-oxoG) on the G base.

Discussion

Here, we used whole genome sequencing of UV-irradiated yeast to elucidate the distinct molecular processes by which UVB and UVA light cause mutations in a model eukaryote. Our data indicate that UVB light induces a diverse spectrum of mutation classes, including C > T, T > C, and C > A substitutions in dipyrimidine sequences and T > A substitutions at TA dinucleotides. All of these mutation classes show transcriptional asymmetry, with fewer mutations on the TS of yeast genes. In contrast, UVA light primarily induces a low frequency of C > A/G > T and C > G/G > C substitutions in WT cells, the former of which are significantly elevated in cells lacking the key BER enzyme Ogg1. UVA mutations in ogg1Δ mutant cells do not show transcriptional asymmetry, but instead exhibit replication asymmetry, with UVA-induced G > T substitutions being enriched on the leading strand of the replication fork. Finally, using a new mutation reporter system, we show that UVA-induced 8-oxoG lesions can induce the C > A/G > T mutation class responsible for the NRAS Q61K driver mutation in melanoma.

Our analysis of mutations arising from yeast cells repeatedly exposed to a sublethal dose of UVB irradiation revealed a surprisingly diverse mutation spectrum composed of not only UV signature C > T substitutions but also C > A and T > C substitutions in dipyrimidine sequences and T > A substitutions in a TA sequence context. This mutation spectrum is very similar to our previously published data for UVC-exposed yeast (Laughery et al. 2020). Notably, we observed more C > T mutations in UVB-exposed yeast, consistent with prior UV mutagenesis studies (Sage et al. 1996; Kozmin et al. 2003) and reports indicating that UVB favors cytosine-containing photoproducts more than UVC irradiation (Mitchell et al. 1992). Analysis of mutation frequency and transcriptional asymmetry in both WT and repair-deficient (e.g. rad16Δ) yeast strains revealed that while the C > A and T > C substitutions are likely due to canonical UV photoproducts occurring at dipyrimidine sequences (e.g. CPD or 6-4PP), the T > A substitutions may primarily be caused by an atypical TA photoproduct. Similar mutation classes (e.g. C > A, T > C, and T > A) were also observed upon genome sequencing of UV-exposed mammalian cells (e.g. Kucab et al. 2019); however, the relatively high frequency of background mutations in mammalian cells grown in cell culture made it unclear whether these mutations were caused by UV exposure.

While UVB exposure causes mutations by inducing helix-distorting photoproducts that are repaired by NER, our data indicate that repeated UVA exposure primarily generates guanine base lesions that are repaired by BER. In BER-competent yeast, the frequency of UVA-induced mutations is orders of magnitude lower than that of equivalent doses of UVB or UVC. However, we show that deletion of the gene encoding the 8-oxoG glycosylase Ogg1 results in a synergistic increase in UVA-induced mutations in yeast, consistent with a previous report (Kozmin et al. 2005). Previous studies have indicated that UVA can yield singlet oxygen species when absorbed by endogenous cellular photosensitizers, such as riboflavin, FMN, and FAD (Baier et al. 2006). Since singlet oxygen preferentially damages guanine bases to form 8-oxoG lesions (Devasagayam et al. 1991; Sies and Menck 1992), this can explain the preponderance of C > A/G > T substitutions in UVA-irradiated ogg1 cells in our data and in previous reports analyzing yeast WT and repair-deficient yeast exposed to UVA (Kozmin et al. 2005). In contrast, exposure of yeast to hydrogen peroxide induces a distinct mutation spectra consisting primarily of C > T substitutions, which likely arise from the oxidation of cytosine bases (Kozmin et al. 2005; Degtyareva et al. 2013; Degtyareva et al. 2019). The differences in these mutation spectra likely reflect the relative abundance of different classes of reactive oxygen species (e.g. singlet oxygen vs hydroxyl radical) arising from UVA or hydrogen peroxide exposure.

The mutation spectra of UVA-irradiated ogg1Δ cells revealed a striking enrichment of G > T mutations in contexts where the guanine base is flanked by thymine. For example, ∼40% of G > T mutations occur in a TGT sequence context, and nearly 90% occur in a TGN or NGT sequence context. This genome-wide mutation spectrum is consistent with a reported can1 mutation spectrum from UVA-irradiated yeast, which had mutation hotspots at ACA/TGT sequences (Kozmin et al. 2005), suggesting that these hotspots are a function of the DNA damaging agent, not phenotypic selection. A possible mechanism for this recurrent mutation pattern could be elevated 8-oxoG lesion formation at TGT sequences. However, genome-wide profiling of 8-oxoG lesions in yeast indicated that TGT (and TGN and NGT) sequences have lower levels of 8-oxoG formation than other sequence contexts (Wu et al. 2018). It is also possible that this sequence pattern arises from UVA-specific preferences in 8-oxoG formation, although the observation that non-UVA–exposed ogg1Δ cells have a somewhat similar mutation spectrum does not support this hypothesis. An alternative explanation could be that DNA polymerase bypass of the UVA-induced 8-oxoG lesion uses a neighboring base (T) to template the misinsertion opposite the lesion. This replication slippage model is known to contribute to DNA polymerase errors (Kunkel and Soni 1988) and may potentially explain the high frequency of G > T mutations in TGN and NGT sequence contexts. A similar mechanism may potentially explain the observed substitution patterns at NTT sequence contexts in UVB-exposed yeast cells (Supplementary Fig. 6). Our previous report indicates that UVC exposure, like UVA, can also induce G > T mutations in a TGT sequence context (i.e. C > A in ACA context), albeit at low frequency (Laughery et al. 2020). Since this mutation type is responsible for the oncogenic NRAS Q61K substitution observed in melanoma (Hodis et al. 2012; Forbes et al. 2017), it will be important in future studies to determine to what extent it is induced by solar UV exposure in human skin cells.

Unlike UVB-induced mutations, UVA-induced C > A/G > T substitutions do not show transcriptional asymmetry in BER-deficient yeast. These findings suggest that in yeast, UVA-induced 8-oxoG lesions are not subject to transcription-coupled repair. Consistent with this finding, previous studies have indicated that 8-oxoG does not cause a significant block to RNA polymerase transcription (Kuraoka et al. 2003; Kathe et al. 2004; Tornaletti et al. 2004; Wang et al. 2018b; Chakraborty et al. 2021), which is normally a prerequisite for initiating TC-NER. Previous analysis of 8-oxoG repair in human cells revealed preferential repair of the TS of the ATM gene; however, this was dependent on the hOGG1 glycosylase (Guo et al. 2013; Wang et al. 2018b). This suggests that the TC-NER substrate may be the glycosylase-induced abasic site, which is known to stall RNA polymerase and be repaired by TC-NER (Tornaletti et al. 2006; Kim and Jinks-Robertson 2010; Wang et al. 2018a; Thompson and Cortez 2020), rather than the intact 8-oxoG lesion itself. Further guanine oxidation products such as spiroiminodihydantoin and guanidinohydantoin are also known to be repaired by NER (Shafirovich et al. 2016; Chakraborty et al. 2021); however, it is possible that these lesions are not significantly induced by UVA exposure.

Our analysis indicates that UVA-induced G > T mutations in BER-deficient cells (i.e. ogg1Δ) are enriched on the leading strand during DNA replication, while UVB-induced C > T mutations do not show significant replication asymmetry. Previous studies have indicated that G > T mutations arising from endogenous 8-oxoG lesions in ogg1Δ mutant yeast are also enriched on the leading strand (Pavlov et al. 2002), consistent with our results. Two mechanisms may contribute to this replication strand asymmetry: first, mismatch repair has been reported to remove adenine misinsertion events opposite 8-oxoG more efficiently from the lagging strand (Pavlov et al. 2003). Second, DNA polymerase eta (Rad30 in yeast) also suppresses 8-oxoG–induced G > T mutations preferentially on the lagging strand (Mudrak et al. 2009). It will be important to determine whether these or other mechanisms contribute to the replication asymmetry that we observed for UVA-induced mutations in yeast. Since it has been reported that the human OGG1 glycosylase is frequently downregulated in basal cell carcinomas (Huang et al. 2012), it will be interesting to determine whether exposure to the UVA spectrum of sunlight causes similar mutation classes in human skin cancers and whether these are enriched on the leading strand during DNA replication.

In summary, our results in our results indicate that UVB and UVA light induce many noncanonical UV mutations (e.g. T > C, T > A, and C > A substitutions) and provide insight into the underlying mutagenic mechanisms. Since many of these mutation classes are responsible for oncogenic mutations in key driver genes in melanoma (e.g. NRAS Q61K and NRAS Q61R), our findings may have important implications for UV mutagenesis in human skin cancers.

Data availability

Strains and plasmids are available upon request. The whole genome sequencing data used to identify yeast UVA and UVB mutations have been submitted to the NCBI Sequence Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) under BioProject accession number SRA: PRJNA888347. Code used to analyze mutation data is available at https://github.com/bmorledge-hampton19/UVA-and-UVB-Mutagenesis-in-Yeast.

Supplemental material available at GENETICS online.

Funding

This research was supported by the following grants from NIEHS: R21ES029655 (J.J.W.), R01ES028698 (J.J.W.), and R01ES032814 (S.A.R. and J.J.W.).